Multi-language-based neural machine translation model

Abstract

The invention discloses a multi-language-based neural machine translation model, and relates to the field of neural machine translation, and the multi-language-based neural machine translation model comprises a multi-language data set construction process, a multi-language translation system and a multi-language translation model operation process. A multi-language parallel corpus and a multi-language neural machine translation model are manually constructed; the multi-language parallel corpus construction method mainly constructs multi-language parallel corpora of Russian-Uzburk-Uygur-English-Chinese, by means of existing bilingual data, other three languages are obtained through translation by means of translation tools such as xiaoniu and google, and the multi-language parallel corpora are constructed through similarity calculation and manual screening. The multi-language neural machine translation model is constructed on the basis of a transform framework, and the constructed data set is used for training the multi-language translation model.

Description

technical field [0001] The invention relates to the field of neural machine translation, in particular to a multilingual-based neural machine translation model. Background technique [0002] Nowadays, there are artificially constructed multilingual parallel corpora and multilingual neural machine translation models; the former mainly constructs Russian-Uzbek-Uyghur-English-Chinese multilingual parallel corpus, with the help of existing bilingual data, using Mavericks and Google, etc. The translation tool translates the other three languages, and constructs multilingual parallel corpora through calculation of similarity and manual screening; the latter builds a multilingual neural machine translation model based on the transformer framework, and uses the constructed data set to train the multilingual translation model. The multilingual corpus in the network is very rare, and most of the multilingual parallel corpus has the problem of data imbalance, and the language correlati...

AI Technical Summary

Problems solved by technology

The multilingual corpus in the network is very rare, and most of the multilingual parallel corpus has the problem of data imbalance, and the language correlation in the multilingual parallel corpus is not strong, so the Russian-Russian-Uzbek-Uyghur language - English-Chinese multilingual parallel corpus, the similarity between the five languages ​​is high and the amount of data is equal

Images